Perfect matchings for the three-term Gale-Robinson sequences Mireille Bousquet-Mélou, James Propp, Julian West To cite this version: Mireille Bousquet-Mélou, James Propp, Julian West. Perfect matchings for the three-term Gale- Robinson sequences. The Electronic Journal of Combinatorics, Open Journal Systems, 2009, 16 (1), paper R125. hal-00396223 HAL Id: hal-00396223 https://hal.archives-ouvertes.fr/hal-00396223 Submitted on 17 Jun 2009 HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non, lished or not. The documents may come from émanant des établissements d’enseignement et de teaching and research institutions in France or recherche français ou étrangers, des laboratoires abroad, or from public or private research centers. publics ou privés. PERFECT MATCHINGS FOR THE THREE-TERM GALE-ROBINSON SEQUENCES MIREILLE BOUSQUET-MELOU,´ JAMES PROPP, AND JULIAN WEST Abstract. In 1991, David Gale and Raphael Robinson, building on explorations carried out by Michael Somos in the 1980s, introduced a three-parameter family of rational recurrence relations, each of which (with suitable initial conditions) appeared to give rise to a sequence of integers, even though a priori the recurrence might produce non-integral rational numbers. Throughout the ’90s, proofs of integrality were known only for individual special cases. In the early ’00s, Sergey Fomin and Andrei Zelevinsky proved Gale and Robinson’s integrality conjecture. They actually proved much more, and in particular, that certain bivariate ratio- nal functions that generalize Gale-Robinson numbers are actually polynomials with integer coefficients. However, their proof did not offer any enumerative interpretation of the Gale- Robinson numbers/polynomials. Here we provide such an interpretation in the setting of perfect matchings of graphs, which makes integrality/polynomiality obvious. Moreover, this interpretation implies that the coefficients of the Gale-Robinson polynomials are positive, as Fomin and Zelevinsky conjectured. In memory of David Gale, 1921-2008 1. Introduction Linear recurrences are ubiquitous in combinatorics, as part of a broad general framework that is well-studied and well-understood; in particular, many combinatorially-defined sequences can be seen on general principles to satisfy linear recurrences (see [26]), and conversely, when an integer sequence is known to satisfy a linear recurrence it is often possible to reverse-engineer a combinatorial interpretation for the sequence (see [4] and references therein for a general discussion, and [3, Chapter 3] for specific examples). In contrast, rational recurrences such as s(n) = (s(n − 1)s(n − 3) + s(n − 2)2)/s(n − 4), which we prefer to write in the form s(n)s(n − 4) = s(n − 1)s(n − 3) + s(n − 2)2, are encountered far less often, and there is no simple general theory that describes the solutions to such recurrences or relates those solutions to combinatorial structures. The particular rational recurrence relation given above is the Somos-4 recurrence, and is part of a general family of recurrences introduced by Michael Somos: s(n)s(n − k)= s(n − 1)s(n − k +1)+ s(n − 2)s(n − k +2)+ · · · + s(n −⌊k/2⌋)s(n −⌈k/2⌉). If one puts s(0) = s(1) = · · · = s(k − 1) = 1 and defines subsequent terms using the Somos- k recurrence, then one gets a sequence of rational numbers which for the values k = 4, 5, 6, 7 is actually a sequence of integers. (Sequences Somos-4 through Somos-7 are entries A006720 through A006723 in [24].) Although integer sequences satisfying such recurrences have received a fair bit of attention in the past few years, until recently algebra remained one step ahead of combinatorics, and there was no enumerative interpretation of these integer sequences. (For links related to Somos sequences, see http://jamespropp.org/somos.html.) Date: 16 June 2009. JP was supported by grants from the National Security Agency and the National Science Foundation. JW was supported by the National Sciences and Engineering Research Council of Canada. 1 2 MIREILLE BOUSQUET-MELOU,´ JAMES PROPP, AND JULIAN WEST Inspired by the work of Somos, David Gale and Raphael Robinson [13, 12] considered se- quences given by recurrences of the form a(n)a(n − m)= a(n − i)a(n − j)+ a(n − k)a(n − ℓ), with initial conditions a(0) = a(1) = · · · = a(m − 1) = 1, where m = i + j = k + ℓ. We call this the three-term Gale-Robinson recurrence 1. The Somos-4 and Somos-5 recurrences are the special cases where (i,j,k,ℓ) is equal to (3, 1, 2, 2) and (4, 1, 3, 2) respectively. Gale and Robinson conjectured that for all integers i,j,k,ℓ > 0 with i + j = k + ℓ = m, the sequence a(0),a(1),... determined by this recurrence has all its terms given by integers. About ten years later, this was proved algebraically in an influential paper by Fomin and Zelevinsky [11]. 1.1. Contents In this paper, we first give a combinatorial proof of the integrality of the three-term Gale- Robinson sequences. The integrality comes as a side-effect of producing a combinatorial in- terpretation of those sequences. Specifically, we construct a sequence of graphs P (n; i,j,k,ℓ) (n ≥ 0) and prove in Theorem 9 that the nth graph in the sequence has a(n) (perfect) match- ings. Our graphs, which we call pinecones, generalize the well-known Aztec diamond graphs, which are the matchings graphs for the Gale-Robinson sequence1, 1, 2, 8, 64,1024,... in which i = j = k = ℓ = 1. A more generic example of a pinecone is shown in Figure 1. All pinecones are subgraphs of the square grid. Figure 1. The pinecone P (25;6, 2, 5, 3). Its matching number is a(25), where a(n) is the Gale-Robinson sequence associated with (i,j,k,ℓ)=(6, 2, 5, 3). We give two ways to construct pinecones for the Gale-Robinson sequences: a recursive method (see Figure 11 and the surrounding text) that constructs the graph P (n; i,j,k,ℓ) in terms of the smaller graphs P (n′; i,j,k,ℓ) with n′ <n, and a direct method (see Formula (2) in Section 3) that allows one to construct the graph P (n; i,j,k,ℓ) immediately. The heart of our proof is the demon- stration that if one defines a(n) as the number of perfect matchings of P (n) ≡ P (n; i,j,k,ℓ), the sequence a(0),a(1),a(2), ... satisfies the Gale-Robinson recurrence. This fact, in combina- tion with a simple check that a(0) = a(1) = · · · = a(m − 1) = 1, gives an immediate inductive validation of our claim that P (n) has a(n) perfect matchings for all n, which yields additionally the integrality of a(n). General pinecones are defined in Section 2, where we also explain how to compute induc- tively their matching number via Kuo’s condensation lemma [17]. In Section 3, we describe how to associate a sequence of pinecones to a Gale-Robinson sequence, and observe that for these pinecones, the condensation lemma specializes precisely to the Gale-Robinson recurrence. Indeed, the recursive method of constructing pinecones, in combination with Kuo’s condensa- tion lemma, gives combinatorial meaning to the different terms a(n1)a(n2) of the Gale-Robinson recurrence. In Section 4, we refine our argument to prove that the sequence p(n) ≡ p(n; w,z) defined by p(n)p(n − m)= wp(n − i)p(n − j)+ zp(n − k)p(n − ℓ), 1Gale and Robinson also considered recurrences of the form a(n)a(n − m) = a(n − g)a(n − h)+ a(n − i)a(n − j)+ a(n − k)a(n − ℓ) for suitable values of g,h,i,j,k,ℓ,m, but such four-term Gale-Robinson recurrences will not be our main concern here. PERFECT MATCHINGS FOR THE THREE-TERM GALE-ROBINSON SEQUENCES 3 with i + j = k + ℓ = m and p(0) = p(1) = · · · = p(m − 1) = 1, is a sequence of polynomials in w and z with nonnegative integer coefficients. More precisely, we prove in Theorem 20 that p(n; u2, v2) counts perfect matchings of the pinecone P (n; i,j,k,ℓ) by the number of special horizontal edges (the exponent of the variable u) and the number of vertical edges (the exponent of the variable v). The fact that p(n) is a polynomial with coefficients in Z was proved in [11], but no combinatorial explanation was given and the non-negativity of the coefficients was left open. 1.2. Strategy, and connections with previous work For much of the work in this paper, we share precedence with the students in the NSF- funded program REACH (Research Experiences in Algebraic Combinatorics at Harvard), led by James Propp, whose permanent archive is on the web at http://jamespropp.org/reach/. A paper by one of these students, David Speyer [25], introduced a very flexible framework (the “crosses and wrenches method”) that, starting from a recurrence relation of a certain type, constructs a sequence of graphs whose matching numbers satisfy the given recurrence. This framework includes the three-term Gale-Robinson recurrences, and thus yields a combinatorial proof of the integrality of the associated sequences. This extends to a proof that the bivariate Gale-Robinson polynomials mentioned above are indeed polynomials, and have non-negative coefficients. One difference with our paper is that Speyer’s graphs are only described explicitly for Somos-4 and Somos-5 sequences, whereas our construction is explicit for any Gale-Robinson sequence.